How-to Sync Data From JDBC Sources

Intro

This is an end-to-end guide about how to migrate tables from JDBC sources (MySQL, PostgreSQL, etc.) to iomete and display it in the BI dashboard.

📘

First, you need to establish an SSH tunnel between iomete and your database in your private network. See Database Connection Options

---

Database to migrate

Let's assume that we want to replicate the MySQL database (or any other supported JDBC database) to the iomete warehouse

📘

In this tutorial, we will be using a publicly accessible iomete-tutorial database instance that contains the Employees Sample Database.

📘

In case of connecting to your own database instance see Database Connection Options for the details

Here are the details of iomete-tutorial public database:

Host: iomete-tutorial.cetmtjnompsh.eu-central-1.rds.amazonaws.com
Port: 3306
Username: tutorial_user
Password: 9tVDVEKp

The database contains the following tables:

---

Create warehouse

Create a new warehouse instance

---

Querying Source Table

After having the warehouse created, we create a table using JDBC Sources using CREATE TABLE command. In the OPTIONS part we specify credentials of the database to which we want to connect as follows (see JDBC Sources):

CREATE TABLE IF NOT EXISTS employees_proxy
USING org.apache.spark.sql.jdbc
OPTIONS (
  url "jdbc:mysql://iomete-tutorial.cetmtjnompsh.eu-central-1.rds.amazonaws.com:3306/employees",
  dbtable "employees.employees",
  driver 'com.mysql.cj.jdbc.Driver',
  user 'tutorial_user',
  password '9tVDVEKp'
);

SELECT * FROM employees_proxy limit 100;

📘

This table doesn't hold the actual data. Data will be retrieved from the actual source once we query the table

---

Migrating Data

To move the data from the source to the warehouse, you can use one of the following options:

Non-partitioned Table

• Option 1. Create a table from select

-- Create table directly from the query
CREATE TABLE employees 
  AS SELECT  * FROM employees_proxy;
-- To inspect the table use the following query
DESC TABLE EXTENDED employees;

• Option 2. Insert into to existing table

--just append data
INSERT INTO employees
  SELECT  * FROM employees_proxy
--or you can use the follwing command to overwerite data
--first clean an existing data and then insert new data
INSERT OVERWRITE TABLE employees
  SELECT  * FROM employees_proxy

• Option 3. Merge with existing data

MERGE INTO employees
  USING (SELECT  * FROM employees_proxy) updates
  ON employees.emp_no = updates.emp_no
WHEN MATCHED THEN
  UPDATE SET *
WHEN NOT MATCHED
  THEN INSERT *

Partitioning data to speed up queries or DML that have predicates involving the partition columns. Here let's create an artificial column birth_year from birth_date and partition data by birth_year

SELECT SUBSTRING(birth_date, 0, 4) as birth_year, * FROM employees_proxy LIMIT 100;

Partitioned Table

• Option 1. Create a partitioned table from select

-- Create a partitioned table directly from the query
CREATE TABLE employees_partitioned 
  PARTITIONED BY (birth_year)
AS SELECT SUBSTRING(birth_date, 0, 4) as birth_year, * FROM 
      employees_proxy order by birth_year; 

-- To inspect the table use the following query
DESC TABLE EXTENDED employees_partitioned;

• Option 2. Insert into to existing table

--just append data
INSERT INTO employees_partitioned
   SELECT SUBSTRING(birth_date, 0, 4) as birth_year, * FROM employees_proxy order by birth_year;

--or you can use the follwing command to overwerite data

--first clean an existing data and then insert new data
INSERT OVERWRITE TABLE employees_partitioned
   SELECT SUBSTRING(birth_date, 0, 4) as birth_year, * FROM employees_proxy order by birth_year;

• Option 3. Merge with existing data

MERGE INTO employees_partitioned
 USING (SELECT 
   SUBSTRING(birth_date, 0, 4) as birth_year, * FROM employees_proxy) updates
 ON employees_partitioned.emp_no = updates.emp_no
WHEN MATCHED THEN
  UPDATE SET *
WHEN NOT MATCHED
  THEN INSERT *

---

Visualize Data

Let's move employees.salaries before moving to BI visualization

CREATE TABLE IF NOT EXISTS salaries_proxy
USING org.apache.spark.sql.jdbc
OPTIONS (
  url "jdbc:mysql://iomete-tutorial.cetmtjnompsh.eu-central-1.rds.amazonaws.com:3306/employees",
  dbtable "employees.salaries",
  driver 'com.mysql.cj.jdbc.Driver',
  user 'tutorial_user',
  password '9tVDVEKp'
);

CREATE TABLE salaries
  AS SELECT  * FROM salaries_proxy;

Create a view joining employees and salaries tables

CREATE OR REPLACE VIEW employee_salaries AS 
 SELECT e.emp_no, e.first_name, e.last_name, e.gender, s.salary FROM employees e
  JOIN salaries s ON e.emp_no = s.emp_no;

Open BI Application

Add new dataset

To add a new dataset, choose Dataset link from the Data menu and click the +Dataset button:

Create a new chart

Click on the newly created dataset employee_salaries which opens chart view. Let's create a table visualization for the Top 10 High Salary Employees

Save the chart to a dashboard

Create another chart. This time Female/Male Salary Distribution using PieChart visualization

Save this chart to the dashboard too and navigate to the dashboard. And, here is the dashboard of the Employees that we just created

Congratulations 🎉🎉🎉

📘

Bonus part

There is a dedicated python library to help to automate this table replication with just a configuration. Please, check out Syncing JDBC Sources